CEO_OPT TDDFT

yan huang

unread,

Aug 25, 2022, 5:46:32 AM8/25/22

to cp2k

Dear all,

Now I want to get the first excited state structure, should I add the following lines in cp2k 9.1?

&EXCITED_STATES
STATE 1
XC_KERNEL_METHOD BEST_AVAILABLE
&END EXCITED_STATES

Thanks & regards

Best,

Yan

Anna Hehn

unread,

Aug 25, 2022, 7:05:15 AM8/25/22

to cp...@googlegroups.com

Dear Yan,

you should add the EXCITED_STATES section in the DFT section, specifying the state to be optimized, and in the PROPERTIES section you can define the kernel to be used. I would recommend to use an ADMM-approximated hybrid functional kernel for high accuracy.

For more details have a look at the tutorial on the CP2K website, https://www.cp2k.org/howto:tddft. Also you can find further exemplary input files of corresponding publications on the Materials Cloud platform, https://archive.materialscloud.org/record/2022.81.

Let me know in case of further questions,

best regards,

Anna

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/04c2649f-9a63-4d7a-ba69-295901a7482en%40googlegroups.com.

yan huang

unread,

Aug 27, 2022, 12:14:13 AM8/27/22

to cp2k

Dear Anna,

Thank you very much for you reply! I run the example for the CP2K and get the opt-pos-1.xyz file. I want to know if the opt-pos-1.xyz is the excited-state structure?

the inp file、out file and xyz file have been attatced.

best regards,

Yan

tddft.inp

opt-pos-1.xyz

opt.out

Anna Hehn

unread,

Aug 28, 2022, 5:44:39 PM8/28/22

to cp...@googlegroups.com

Dear Yan,

please excuse the late reply. Yes, the structure of the state that is optimized is printed in the xyz file.

Best regards

Anna

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/674eb227-d64e-4b6e-bd40-c83fe4d8e11an%40googlegroups.com.

yan huang

unread,

Sep 3, 2022, 5:52:09 AM9/3/22

to cp2k

Dear Anna,

Thank you very much for your reply! Recently, I calculated the structure of excited states for my system. Because of the large system, I used the PBE functional instead of the hybrid functional. But I found excited stated forces reach the maximun number of iterations, did't converge. I want to konw how to increase the the maximun number of iterations (default : 100)?

the inp file、out file and xyz file have been attatced.

92 PCG F 0.13E-01 0.0000000422 1245.73
93 PCG F 0.39E-01 0.0000000362 1259.35
94 PCG F 0.37E-01 0.0000000446 1272.88
95 PCG F 0.24E-01 0.0000000310 1286.32
96 PCG F 0.47E-01 0.0000000218 1299.98
97 PCG F 0.44E-01 0.0000000422 1313.46
98 PCG F 0.11E-01 0.0000000248 1327.00
99 PCG F 0.40E-01 0.0000000205 1340.44
100 PCG F 0.23E-01 0.0000000206 1353.98
The linear solver didn't converge! Maximum number of iterations reached.
DEBUG:: Pz*dT 0.02698408 0.01755403 0.08408835
DEBUG:: Pz*dHppl -0.00172389 -0.00192144 -0.00758440
DEBUG:: Pz*dHppnl 0.00203508 0.00898434 0.03836144
DEBUG:: Force Pz*dHcore 0.02729528 0.02461693 0.11486539
DEBUG:: Pz*dVhxc -0.05201569 -0.05475358 -0.22575630
DEBUG:: Vh(rhoz)*dncore -0.08077092 -0.03255382 -0.27401921
DEBUG:: Pin*dK*rhoz 0.09644318 0.04774496 0.32286211
DEBUG:: Force Pin*V(rhoz) -0.03634344 -0.03956245 -0.17691341
DEBUG:: Wz*dS -0.00842473 0.00079797 -0.00826806
DEBUG:: Response Force -0.01747288 -0.01414755 -0.07031609
DEBUG:: Total Force -0.01747288 -0.01414755 -0.07031609

Best regards

Yan Huang

Co.xyz

opt.out

opt.inp

Anna Hehn

unread,

Sep 3, 2022, 8:01:07 PM9/3/22

to cp...@googlegroups.com

Dear Yan,

there is a MAX_ITER keyword in the LINRES section of the PROPERTIES section (and also in the TDDFPT section).

The number of max SCF cycles in OUTER_SCF should be chosen as large as needed and in the inner SCF the number should be smaller, so the other way around.

Maybe one could also try different optimizers like DIIS.

Best regards,

Anna

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ad1dd997-c301-4866-979e-93a41e699d98n%40googlegroups.com.

yan huang

unread,

Sep 4, 2022, 4:37:58 AM9/4/22

to cp2k

Dear Anna,

Thank you very much for your reply! I have been added the MAX_ITER keyword (default : 50) in the LINRES section and set the value 300. But it didn't change the maximun number of excited stated forces iterations.

&PROPERTIES
&TDDFPT
KERNEL FULL
NSTATES 5
MAX_ITER 300
CONVERGENCE [eV] 1E-6
RKS_TRIPLETS .F.
&DIPOLE_MOMENTS
DIPOLE_FORM BERRY
&END DIPOLE_MOMENTS
&XC
&XC_FUNCTIONAL PBE
&END XC_FUNCTIONAL
&END XC
&END TDDFPT

&LINRES
MAX_ITER 300
&END LINRES
&END PROPERTIES
&END FORCE_EVAL

87 PCG F 0.40E-01 0.0000000203 1189.10
88 PCG F 0.15E-01 0.0000000073 1202.69
89 PCG F 0.32E-01 0.0000000106 1216.32
90 PCG F 0.30E-01 0.0000000073 1230.02
91 PCG F 0.24E-01 0.0000000145 1243.78
92 PCG F 0.26E-01 0.0000000052 1257.48
93 PCG F 0.81E-01 0.0000000051 1271.12
94 PCG F 0.71E-01 0.0000000069 1284.73
95 PCG F 0.30E-01 0.0000000066 1298.31
96 PCG F 0.38E-01 0.0000000046 1311.90
97 PCG F 0.82E-01 0.0000000042 1325.66
98 PCG F 0.44E-01 0.0000000064 1339.29
99 PCG F 0.25E-01 0.0000000046 1352.89
100 PCG F 0.23E-01 0.0000000068 1366.48

The linear solver didn't converge! Maximum number of iterations reached.

DEBUG:: Pz*dT 0.00157382 0.00048231 0.00736383
DEBUG:: Pz*dHppl -0.00008937 -0.00004994 -0.00067884
DEBUG:: Pz*dHppnl -0.00001782 0.00017519 0.00369994
DEBUG:: Force Pz*dHcore 0.00146663 0.00060756 0.01038493
DEBUG:: Pz*dVhxc -0.00331493 -0.00167409 -0.01979951
DEBUG:: Vh(rhoz)*dncore -0.00513353 -0.00008315 -0.02284556
DEBUG:: Pin*dK*rhoz 0.00716620 0.00070356 0.02667802
DEBUG:: Force Pin*V(rhoz) -0.00128226 -0.00105368 -0.01596705
DEBUG:: Wz*dS -0.00140176 0.00003803 -0.00036498
DEBUG:: Response Force -0.00121739 -0.00040810 -0.00594710
DEBUG:: Total Force -0.00121739 -0.00040810 -0.00594710

Best regards

Yan Huang

opt.inp

opt.out

Anna Hehn

unread,

Sep 6, 2022, 5:25:06 AM9/6/22

to cp...@googlegroups.com

Dear Yan,

yes please excuse, you're right that this is not a default option in the trunk.

Please just add the following 3 lines in src/excited_states.F instead of the already existing line 115:

NULLIFY (linres_section)

linres_section => section_vals_get_subs_vals(qs_env%input,"PROPERTIES%LINRES")

CALL response_equation(qs_env, p_env,ex_env%cpmos,unit_nr,lr_section=linres_section)

also initializing the variable at the beginning of excited_state_energy subroutine:

TYPE(section_vals_type), POINTER :: linres_section

You can also use my branch maxiter which I just generated and download it from GitHub.

However, I would also consider modifying your settings. As far as I know, CENTER_COORDINATES should be used in combination with molecules and not periodic systems. Also you could maybe check convergence of grids for your system etc.

I hope this helps, again please excuse the inconvenience and let me know if I can help with any further questions.

Best reagrds

Anna

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/be5b6b75-f714-46db-b23d-74e1337e7c9dn%40googlegroups.com.

yan huang

unread,

Sep 8, 2022, 5:40:45 AM9/8/22

to cp2k

Dear Anna,

Thank you very much for your reply! I have been added the following 3 lines in src/excited_states.F instead of the already existing line 115:

IF (dft_control%qs_control%semi_empirical) THEN

NULLIFY (linres_section)
linres_section => section_vals_get_subs_vals(qs_env%input,"PROPERTIES%LINRES")
CALL response_equation(qs_env, p_env,ex_env%cpmos,unit_nr,lr_section=linres_section)

ELSEIF (dft_control%qs_control%dftb) THEN

also initializing the variable at the beginning of excited_state_energy subroutine:

TYPE(qs_p_env_type), POINTER :: p_env
TYPE(section_vals_type), POINTER :: linres_section

After recompiling and reinstalling, I still can't fix the problem changing the maximun number of excited stated forces iterations . Can you give me the link to your GitHub?

And is the position of LINRES section right?

&END TDDFPT
&LINRES
MAX_ITER 300
&END LINRES
&END PROPERTIES

Best regards

Yan Huang

Co.xyz

opt.inp

Anna Hehn

unread,

Sep 8, 2022, 6:06:11 AM9/8/22

to cp...@googlegroups.com

Dear Yan,

yes I am sorry, I checked again and I forgot to tell you to initialize the routines. Please excuse the inconvenience.

The link to the branch is: https://github.com/annahehn/cp2k/tree/maxiter. It is more or less the latest master.

diff --git a/src/excited_states.F b/src/excited_states.F

index f06157953..581a029d1 100644

--- a/src/excited_states.F

+++ b/src/excited_states.F

@@ -34,6 +34,8 @@ MODULE excited_states

USE response_solver, ONLY: response_equation,&

response_force,&

response_force_xtb

+ USE input_section_types, ONLY: section_vals_get_subs_vals,&

+ section_vals_type

#include "./base/base_uses.f90"

IMPLICIT NONE

@@ -75,6 +77,7 @@ CONTAINS

TYPE(qs_energy_type), POINTER :: energy

TYPE(qs_force_type), DIMENSION(:), POINTER :: ks_force, lr_force

TYPE(qs_p_env_type) :: p_env

+ TYPE(section_vals_type), POINTER :: linres_section

CALL timeset(routineN, handle)

@@ -107,7 +110,9 @@ CONTAINS

CALL zero_qs_force(lr_force)

CALL set_qs_env(qs_env, force=lr_force)

!

- CALL response_equation(qs_env, p_env, ex_env%cpmos, unit_nr)

+ NULLIFY (linres_section)

+ linres_section => section_vals_get_subs_vals(qs_env%input,"PROPERTIES%LINRES")

+ CALL response_equation(qs_env, p_env,ex_env%cpmos,unit_nr,lr_section=linres_section)

!

CALL get_qs_env(qs_env, dft_control=dft_control)

IF (dft_control%qs_control%semi_empirical) THEN

Let me please know if installation then works and if this helps with converging the computation for your system, if not, I can try to have another look.

Best regards,

Anna

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/89aa80af-d8ac-48f1-9bdc-dfd1c57c89a6n%40googlegroups.com.

yan huang

unread,

Sep 10, 2022, 4:21:15 AM9/10/22

to cp2k

Dear Anna,

Thank you very much for your reply! I have been solved the problem changing the maximun number of excited stated forces iterations . But I found it more difficult to converge, and the value of convergence increase with iterations.The two out files have been attached.

Iteration Method Restart Stepsize Convergence Time
------------------------------------------------------------------------------
1 CG F 0.00E+00 0.0010353714 0.36
2 CG F 0.52E+00 0.0007680304 6.90
3 CG F 0.37E+00 0.0012386086 13.18
4 CG F 0.33E+00 0.0015398605 19.51
5 CG F 0.20E+00 0.0014074471 25.64
6 CG F 0.16E+00 0.0015509455 32.12
7 CG F 0.15E+00 0.0022541178 38.48
8 CG F 0.11E+00 0.0030670538 45.10
9 CG F 0.83E-01 0.0037224131 51.86
10 CG F 0.65E-01 0.0044437804 58.47
11 CG F 0.53E-01 0.0053939725 64.81
12 CG F 0.44E-01 0.0066346043 71.04
13 CG F 0.36E-01 0.0082228215 77.33
14 CG F 0.30E-01 0.0102075903 83.70
15 CG F 0.26E-01 0.0125859742 90.71
16 CG F 0.22E-01 0.0152855073 97.42
17 CG F 0.18E-01 0.0181872107 103.65
18 CG F 0.16E-01 0.0211694290 110.23
19 CG F 0.14E-01 0.0241439294 116.45
20 CG F 0.12E-01 0.0270690365 122.81
21 CG F 0.10E-01 0.0299429363 129.07
22 CG F 0.92E-02 0.0327885753 135.68
23 CG F 0.83E-02 0.0356394345 142.06
24 CG F 0.74E-02 0.0385300372 148.71
25 CG F 0.67E-02 0.0414911888 155.12
26 CG F 0.61E-02 0.0445484675 161.36
27 CG F 0.56E-02 0.0477224413 167.55
28 CG F 0.52E-02 0.0510295412 174.13
29 CG F 0.48E-02 0.0544829974 180.44
30 CG F 0.44E-02 0.0580935825 186.76
31 CG F 0.41E-02 0.0618701028 193.16
32 CG F 0.38E-02 0.0658196763 199.75
33 CG F 0.36E-02 0.0699478695 206.10
34 CG F 0.34E-02 0.0742587654 212.47
35 CG F 0.32E-02 0.0787550188 218.85
36 CG F 0.30E-02 0.0834379381 225.58

Best regards

Yan Huang

opt-old.out

opt-new.out

Reply all

Reply to author

Forward